BRCA1 and BRCA2 Gene Mutations and Colorectal Cancer Risk

Systematic Review and Meta-analysis

Mok Oh; Ali McBride; Seongseok Yun; Sandipan Bhattacharjee; Marion Slack; Jennifer R. Martin; Joanne Jeter; Ivo Abraham


J Natl Cancer Inst. 2018;110(11):1178-1189. 

In This Article

Materials and Methods

We followed the guidelines summarized in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)[18] and the Meta-analysis of Observational Studies in Epidemiology (MOOSE)[19] statements. The full checklists summarizing compliance with both PRISMA and MOOSE are included in the Supplementary Materials (available online).

Study Selection

The following modified participants, interventions, comparators, outcomes, and studies approach (PICOS) guided eligibility screening of studies for inclusion in our study: 1) Participants: human adult participants (older than age 18 years) who tested positive for either BRCA1 or BRCA2 mutations or had a familial risk of BRCA mutation; 2) Intervention: not applicable; 3) Comparisons: cancer-affected and cancer-unaffected BRCA mutation carriers vs population-based incidence rates, and BRCA-negative cancer-affected patients; 4) Outcomes: incidence or prevalence of colorectal cancer, including colon cancer if differentiated; 5) Studies: investigations reporting colorectal cancer incidence data or sufficient data to calculate risk. Excluded were editorials, letters, commentaries, and review papers; publications reporting diseases other than colorectal cancer, such as Lynch syndrome, familial adenomatous polyposis, Peutz-Jeghers syndrome, or Cowden syndrome; and studies without colorectal cancer incidence data or with insufficient data to calculate risk.

Data Sources

A professional health science librarian (JM) supported the search of the following bibliographic databases: PubMed/MEDLINE (1946–2017), Embase (1947–2017), Cochrane Library (1898–2017), and ProQuest Dissertations & Theses (1861–2017). The search was executed by two of the investigators (MO, SY) and included a combination of indexing terms (MeSH terms in PubMed and EMTREE terms in Embase) as well as keyword terms including "Genes, BRCA1," "Genes, BRCA2," and "Colorectal Neoplasms" and translated for each database. We also conducted a manual check of the reference list of key articles for the recent relevant publications through Scopus (1823–2017). The search strategy is included as Supplementary Table 1, available online.

Data Extraction

Two reviewers (MO, IA) independently screened article titles and abstracts for eligibility. Disagreements were resolved through discussion and consensus, with unresolved issues escalated to a third person (AM). The following information was retrieved from each study using a predeveloped worksheet: publication date, location, study design, type of colorectal cancer, description of cases and controls (eg, number, recruitment method, matching, etc.), age and sex of cases and controls, and risk estimates with corresponding 95% CIs. Authors of each trial were contacted for additional information if needed.

Quality Assessment

Quality assessment was performed on each included study by two reviewers (MO, IA) independently using the Newcastle-Ottawa Scale (NOS). The NOS consists of eight items focused on three domains: selection of study groups, ascertainment of the exposure and outcome, and comparability of groups to assess the quality of observational studies. Ratings were based on a star system and studies with a maximum rating of nine. Studies with one to three stars were categorized as low quality, four to six stars categorized as moderate quality, and seven to nine stars categorized as high quality.

Summary Measures

To quantify the risk of colorectal cancer in the meta-analysis, we used the unadjusted odds ratio as the common metric for studies that reported an estimate such as relative risk, odds ratio, standardized incidence ratio (SIR), hazard ratio (HR), or provided sufficient information to compute the risk estimate. The unadjusted (crude) odds ratio (OR) for each study was calculated from a 2 × 2 contingency table created for each study. This calculated OR was used in the meta-analysis including all subgroup meta-analyses. The cohort studies that reported the SIR and case–control studies that adjusted for age and sex were used in subgroup meta-analyses to provide age- and sex-adjusted estimates.

Statistical Analyses

Meta-analytic calculations were performed using Comprehensive Meta-Analysis (v. 3.0). Study-specific relative log OR were weighted by the inverse of their variance to calculate a summary estimate and the corresponding 95% CIs. Assuming varying effect sizes across studies, associations between BRCA gene mutational status and OR was evaluated using DerSimonian and Laird random-effect models. The Cochran Qstatistic was used to measure heterogeneity through a weighted sum of squares; and the I2 statistic [100% × (Qdf)/Q] to quantify the total percentage of variation across each study due to heterogeneity. A P value lower than .05 for the Cochran Q test and I2exceeding 50% were used as a cutoff for statistically significant heterogeneity.[20] Publication bias was presented by using a contour-enhanced funnel plot of standard error against the effect estimate. The Egger linear regression test method was applied to evaluate asymmetry of the funnel plot and a statistically significant publication bias was considered when the P value was less than .10. Due to the limited number of studies in the overall and subgroup meta-analyses, statistical tests comparing the summary effect measures (odds and risk ratio) was not performed,

Subgroup Analyses

Due to the expected heterogeneity, prespecified subgroup meta-analyses were conducted as shown in Figure 1. After stratifying by BRCA1 and BRCA2, we explored study design features (eg, ascertained vs inferred genotyping; colorectal cancer vs colon cancer; age–sex adjusted vs crude) as sources of heterogeneity.

Figure 1.

Flow diagram of the subgroup meta-analyses.